feat(shenron): Integrate C-SHENRON physics-based radar simulation

- Transitions simulation to Model-Based Development (MBD) architecture.
- Added semantic LiDAR sensor with high-fidelity point cloud recording.
- Fixed '0 points' bug by implementing bit-casting for uint32 semantic tags (Tag 10 -> Metal).
- Corrected antenna count hardcoding and range-slicing errors in radar processor.
- Synchronized hardware configs between project settings and simulation engine.
- Automated synthesis and MCAP serialization in src/main.py cleanup phase.
- Verified rich 5-field point cloud output (x, y, z, velocity, magnitude) in showcase.

scripts/ISOLATE/e2e_agent_sem_lidar2shenron_package/lidar.py

@@ -19,15 +19,39 @@ def map_carla_semantic_lidar_latest(carla_sem_lidar_data):
     '''
     Function to map material column in the collected carla ray_cast_shenron to shenron input
     '''
-    # CARLA 0.9.16 [x, y, z, intensity, cos, obj, tag]
-    # We want [x, y, z, tag]
-    carla_sem_lidar_data_crop = carla_sem_lidar_data[:, (0, 1, 2, 6)]
-    temp_list = np.array([0, 4, 2, 0, 11, 5, 0, 0, 1, 8, 12, 3, 7, 10, 0, 1, 0, 12, 6, 0, 0, 0, 0])
-
-    col = temp_list[(carla_sem_lidar_data_crop[:, 3].astype(int))]
-    carla_sem_lidar_data_crop[:, 3] = col
-
-    return carla_sem_lidar_data_crop
+    # CARLA 0.9.16 raw bits for [x, y, z, cos, obj, tag]
+    # We want [x, y, z, legacy_tag]
+    # Tag is at index 6 if padded, or index 5 in raw
+    tag_col_idx = 6 if carla_sem_lidar_data.shape[1] == 7 else 5
+    
+    # CRITICAL: Recover real integer tags from the float32 bitstream
+    # Force float32 cast before view to ensure 1:1 element mapping
+    tags = carla_sem_lidar_data[:, tag_col_idx].astype(np.float32).view(np.uint32)
+    
+    # CARLA 0.9.16 -> Legacy team_code mapping (used by new_map_material)
+    # 0.9.16: 0:none, 1:building, 2:fence, 4:pedestrian, 10:vehicle, 11:wall, 12:traffic_sign, 13:sky, 14:ground, 15:bridge, 16:rail, 17:guardrail, 18:traffic_light, 19:static, 20:dynamic, 21:water, 22:terrain
+    # Legacy: 13:car, 11:pedestrian, 4:fence, 5:pole, 7:traffic_sign, 2:building, 3:wall, 1:sidewalk, 8:vegetation
+    
+    # We create a 256-long lookup table for all possible tags
+    mapping = np.zeros(256, dtype=int)
+    mapping[1] = 2   # Building
+    mapping[2] = 4   # Fence
+    mapping[4] = 11  # Pedestrian
+    mapping[5] = 5   # Pole
+    mapping[8] = 1   # Sidewalk
+    mapping[9] = 8   # Vegetation
+    mapping[10] = 13 # Vehicle
+    mapping[11] = 3  # Wall
+    mapping[12] = 7  # Traffic Sign
+    mapping[18] = 6  # Traffic Light
+    
+    legacy_tags = mapping[tags % 256] # bitwise safety
+    
+    # Return [x, y, z, legacy_tag]
+    res = np.zeros((carla_sem_lidar_data.shape[0], 4))
+    res[:, 0:3] = carla_sem_lidar_data[:, 0:3]
+    res[:, 3] = legacy_tags
+    return res
 
 # def map_carla_semantic_lidar(carla_sem_lidar_data):
 #     '''

scripts/data_to_mcap.py

@@ -191,7 +191,7 @@ def convert_folder(folder_path):
                     writer.add_message(radar_channel_id, log_time=ts_ns, data=json.dumps(radar_msg).encode(), publish_time=ts_ns)
 
             # SHENRON RADAR
-            shenron_file = f"frame_{self.frame_id:06d}.npy"
+            shenron_file = f"frame_{int(frame['frame_id']):06d}.npy"
             shenron_path = os.path.join(folder_path, "shenron_radar", shenron_file)
             if os.path.exists(shenron_path):
                 s_data = np.load(shenron_path)

scripts/generate_shenron.py

@@ -49,7 +49,7 @@ def process_session(session_path):
             if data.shape[1] == 6:
                 # Pad with a dummy intensity column at index 3
                 # This aligns 'tag' to index 6 as expected by our lidar.py mapping
-                padded_data = np.zeros((data.shape[0], 7))
+                padded_data = np.zeros((data.shape[0], 7), dtype=np.float32)
                 padded_data[:, 0:3] = data[:, 0:3] # x, y, z
                 padded_data[:, 4:7] = data[:, 3:6] # cos, obj, tag
                 data = padded_data

scripts/verify_tags.py

@@ -0,0 +1,29 @@
+import numpy as np
+import os
+
+# Path to the failed session's lidar frame
+frame_path = r'd:\CARLA\CARLA_0.9.16\PythonAPI\Fox\data\showcase_20260401_194412\lidar\frame_000001.npy'
+
+if not os.path.exists(frame_path):
+    print(f"File not found: {frame_path}")
+else:
+    data = np.load(frame_path)
+    print(f"Data shape: {data.shape}")
+    
+    # Take the Tag column (index 5)
+    tags_float = data[:, 5]
+    
+    # Method 1: Simple cast (the wrong way)
+    tags_int_wrong = tags_float.astype(int)
+    print(f"Unique tags (wrong way): {np.unique(tags_int_wrong)}")
+    
+    # Method 2: Bit-casting (the right way)
+    # We must ensure it's a copy and contiguous for .view() to work as intended on a slice
+    tags_int_right = tags_float.copy().view(np.uint32)
+    print(f"Unique tags (right way): {np.unique(tags_int_right)}")
+    
+    # Method 3: Bit-casting on the whole row (even safer)
+    # CARLA [x:f32, y:f32, z:f32, cos:f32, obj:u32, tag:u32]
+    # If we read it all as f32, we can recover correctly
+    obj_ids = data[:, 4].copy().view(np.uint32)
+    print(f"Unique object IDs (right way): {len(np.unique(obj_ids))} unique IDs found.")

src/main.py

@@ -23,6 +23,8 @@ sys.path.append(os.path.dirname(os.path.dirname(__file__)))
 import config
 from scenario_loader import load_scenario, list_scenarios
 from scripts.data_to_mcap import convert_folder
+from scripts.generate_shenron import process_session as generate_shenron_data
+from pathlib import Path
 
 
 # -----------------------------------------------------------------------
@@ -281,8 +283,13 @@ def main():
         ego_vehicle.destroy()
         if recorder:
             recorder.close()
-            # AUTO-MCAP Step
+            
             if os.path.exists(recorder.base_path):
+                # NEW: AUTO-SHENRON Step
+                print(f"[AUTO-SHENRON] Synthesizing physics-based radar for: {recorder.base_path}")
+                generate_shenron_data(Path(recorder.base_path))
+
+                # AUTO-MCAP Step
                 print(f"[AUTO-MCAP] Triggering seamless conversion for: {recorder.base_path}")
                 convert_folder(recorder.base_path)